The influence of risk-taking on bank efficiency

THE INFLUENCE OF RISK-TAKING
ON BANK EFFICIENCY: EVIDENCE
FROM COLOMBIA
Miguel Sarmiento and Jorge E. Galán
Documentos de Trabajo
N.º 1537
2015
THE INFLUENCE OF RISK-TAKING ON BANK EFFICIENCY:
EVIDENCE FROM COLOMBIA
THE INFLUENCE OF RISK-TAKING ON BANK EFFICIENCY:
EVIDENCE FROM COLOMBIA (*)
Miguel Sarmiento (**)
BANCO DE LA REPÚBLICA, COLOMBIA AND TILBURG UNIVERSITY
Jorge E. Galán (***)
BANCO DE ESPAÑA
(*) The opinions and statements in this article are the sole responsibility of the authors and do not necessarily
coincide with those of Banco de la República, Banco de España or the Eurosystem.
(**) Email: [email protected]
(***) Corresponding author. Address: Banco de España, Alcalá 48 - 28014 Madrid, Spain. Phone: +34913384221.
Email: [email protected]
Documentos de Trabajo. N.º 1537
2015
The Working Paper Series seeks to disseminate original research in economics and finance. All papers
have been anonymously refereed. By publishing these papers, the Banco de España aims to contribute
to economic analysis and, in particular, to knowledge of the Spanish economy and its international
environment.
The opinions and analyses in the Working Paper Series are the responsibility of the authors and, therefore,
do not necessarily coincide with those of the Banco de España or the Eurosystem.
The Banco de España disseminates its main reports and most of its publications via the Internet at the
following website: http://www.bde.es.
Reproduction for educational and non-commercial purposes is permitted provided that the source is
acknowledged.
© BANCO DE ESPAÑA, Madrid, 2015
ISSN: 1579-8666 (on line)
Abstract
This paper presents a stochastic frontier model with random inefficiency parameters which
captures the influence of risk-taking on bank efficiency and distinguishes the effects among
banks with different characteristics. The model is fitted to a 10-year sample of Colombian
banks. Cost and profit efficiency are found to be over and underestimated, respectively,
when risk measures are omitted or are not accurately modelled. Moreover, the magnitudes
at which similar levels of risk affect bank efficiency vary with size and affiliation. In particular,
domestic and small Colombian banks benefit more from being highly capitalised, while large
and foreign banks benefit from higher exposure to credit and market risk. Holding more
liquid assets is found to affect efficiency only at domestic banks. Lastly, we identify some
channels that can explain these differences and provide insights for prudential regulation.
Keywords: bank efficiency, Bayesian inference, heterogeneity, random parameters, risktaking, stochastic frontier models.
JEL classification: C11, C23, C51, D24, G21, G32.
Resumen
Este trabajo presenta un modelo de frontera estocástica con parámetros aleatorios en el
componente de ineficiencia aplicado al sector bancario. Este modelo permite capturar
la influencia de las exposiciones de riesgo sobre la eficiencia bancaria y distingue estos
efectos entre bancos con diferentes características. El modelo es ajustado a una muestra
de 10 años de bancos colombianos. Se encuentra que, cuando las medidas de riesgo son
omitidas o no son modeladas adecuadamente, las estimaciones de eficiencia en costos y
en beneficios son sobreestimadas y subestimadas, respectivamente. Adicionalmente, las
magnitudes en las cuales niveles similares de riesgo afectan la eficiencia bancaria varían
de acuerdo con el tamaño y el tipo de propiedad de los bancos. En particular, los bancos
pequeños y nacionales se benefician más de altos niveles de capitalización, mientras que
los bancos grandes y extranjeros se benefician de mayores exposiciones de riesgo de
crédito y de mercado. También se encuentra que una mayor tenencia de liquidez tiene
efectos únicamente sobre la eficiencia de los bancos nacionales. Finalmente, se identifican
los canales que pueden explicar esas diferencias y algunas ideas de regulación prudencial.
Palabras clave: eficiencia bancaria, inferencia bayesiana, heterogeneidad, parámetros
aleatorios, toma de riesgo, modelos de frontera estocástica.
Códigos JEL: C11, C23, C51, D24, G21, G32.
1. Introduction
The modern banking theory shows that banks’ behavior is subject to uncertainty derived from the behavior of borrowers, depositors and financial markets
in which banks interact. This type of uncertainty is commonly referred as bank
risk-taking. That is, the amount of risk that banks are willing to tolerate, which depends on competition, regulation and corporate governance (Boyd and De Nicoló,
2005; Laeven and Levine, 2009; Wagner, 2010; Agoraki et al., 2011; Anginer et al.,
2013). In their pursuit of better performance banks tend to engage in more risktaking. However, excessive risk-taking lead the financial system to be highly vulnerable to shocks (Rajan, 2006). During the global financial crisis of 2007-08
excessive bank risk-taking was associated with banking runs, fire-sales, and financial fragility (Brunnermeier and Pedersen, 2009; Shleifer and Vishny, 2010). In
response to this behavior, banking regulators have imposed higher capital and liquidity requirements, leverage ratios, and countercyclical provisions for loan losses,
among other regulatory measures (see Basel III standards in BIS, 2010, 2011).
These measures are intended to discourage risk-taking by imposing higher costs to
banks from assuming more risk. Thus, understanding how risk-taking and regulation influences bank performance is an important concern.
In the banking literature recent studies have addressed these effects using different approaches. Several studies accounting for regulatory effects have found that
stringency of capital regulation is associated with higher bank efficiency, while
limiting banking activities discourages efficiency (Chortareas et al., 2012; Barth
et al., 2013; Berger and Bowman, 2013). Other studies have focused on identifying the relationship between credit risk, capitalization and bank efficiency (see
the seminal work of Berger and DeYoung, 1997). Most of studies exploring these
relationships have found that highly capitalized banks are more cost efficient than
banks with low capitalization levels (Williams, 2004; Altunbas et al., 2007; Lepetit
et al., 2008; Fiordelisi et al., 2011). Furthermore, banks with low cost efficiency
have been found to exhibit higher proportions of bad loans and to be more prone
to default (see Podpiera and Weill, 2008; Tabak et al., 2011, for some evidence
from emerging economies).
From the structural approach, risk-taking has been identified as a crucial element of the banking production process which should be properly modeled into
efficiency measurement (Hughes et al., 2001). Recent studies under this approach
show that failure to account for risk-taking may lead to biased estimations of bank
efficiency and misleading estimates of scale economies and cost elasticities (Hughes
and Mester, 2013; Koetter, 2008; Malikov et al., 2014).
Another widely used approach in the literature is to incorporate risk measures
into frontier efficiency methods such as stochastic frontier analysis (SFA). However, most of studies modeling the effects of risk on efficiency under this approach
BANCO DE ESPAÑA
7
DOCUMENTO DE TRABAJO N.º 1537
incorporate only proxies for credit risk (usually nonperforming loans) and omit
other important risks faced by banks (e.g. liquidity, market or insolvency risk).
Some recent studies have accounted for alternative risk measures as inefficiency
determinants. Radić et al. (2012) find capital and liquidity risk to have relevant
effects on cost and profit efficiency of investment banks in G-7 countries. Also, Pessarossi and Weill (2015) find that a higher capital ratio had a positive influence on
cost efficiency of Chinese banks during 2004-2009, suggesting that capital buffers
may improve cost efficiency. These studies reveal that accounting for risk-exposure
heterogeneity across banks is relevant when measuring bank efficiency. Moreover,
the omission of heterogeneity related to size and type of ownership has been identified as an important source of biases in the estimations of banks inefficiency (Bos
et al., 2009; Feng and Zhang, 2012; Goddard et al., 2014).
Identifying inefficiency determinants and accounting for heterogeneity is particularly important in the Colombian banking sector given the rapid expansion of
the sector in recent years, the important role of foreign institutions and the several mergers and acquisition (M&A) processes that have been carried out. These
characteristics have increased the differences in terms of size and capital structure across institutions, which could affect banks’ risk-taking behavior and performance. Furthermore, since 2002 several regulatory measures have been implemented by the Colombian regulators in order to enhance loan loss provisions, and
to set adequate capital and liquidity requirements to limit risk-taking. These measures were initially motivated by a profound financial crisis at the end of 1990s that
evidenced the vulnerability of the Colombian banking sector to external shocks.
Previous studies, although failing to control by risk, have found gains in efficiency
of Colombian banks in recent years and have identified that large and foreign banks
are more efficient than their counterparts. In this context, recognizing differences
in the way risk exposures affect different types of banks becomes relevant in order
to get more accurate efficiency estimations and a complete understanding of the
effects of risk and macroprudential regulation on bank performance.
The aim of this paper is to identify the influence of risk-taking on cost and profit
efficiency of banks and to differentiate these effects between banks with different
sizes and affiliations. We contribute to the literature by proposing a stochastic
frontier model with random inefficiency coefficients, which allows us to identify
the influence of unobserved heterogeneity sources related to risk-taking on bank
efficiency. Our approach is close to that in Feng and Zhang (2012); Goddard et al.
(2014) in which random parameters are used in order to account for unobserved
bank heterogeneity. However, in our model random parameters are modeled in the
inefficiency component rather than in the frontier. This allows us to estimate consistently in a single step heterogeneous effects of risk-exposure measures on bank
inefficiency. In particular, we account for the influence of capital, liquidity, credit,
BANCO DE ESPAÑA
8
DOCUMENTO DE TRABAJO N.º 1537
and market risk exposures on cost and profit efficiency and distinguish these effects
by types of banks. The inference of the model is carried out via Bayesian methods
that formally incorporate parameter uncertainty and allows deriving bank-specific
distributions of efficiency and risk random coefficients. The model is estimated
for the Colombian banking sector using bank-level data from 2002 to 2012. This
period covers several regulatory measures that were implemented to limit bank
risk-taking and to promote the foreign entry of banks. The period considered
also allows us to assess the effects of the global financial crisis on the efficiency of
Colombian banks.
In line with recent evidence, our findings remark the importance of accounting
for size, affiliation and risk exposure in the estimation of bank efficiency. We find
that cost and profit efficiency are over- underestimated when risk measures are
not accurately modeled (see Hughes et al., 2001; Koetter, 2008; Radić et al., 2012,
for similar results). Furthermore, we identify that size and foreign ownership are
not only important determinants of efficiency but also key characteristics defining
the way credit and market risk, liquidity and capitalization levels affect cost and
profit efficiency. Domestic and small Colombian banks benefit more from being
highly capitalized, while large and foreign banks from engaging in higher credit
and market risk. We find that large Colombian banks exhibit higher efficiency
than small institutions and that foreign and small banks were more affected by
the financial crisis and the regulatory measures introduced after 2008. We explain
the main channels supporting these differences in efficiency among banks with
different characteristics.
2. The Colombian banking sector: performance and regulation
During early 1990s the Colombian banking sector was gradually introduced
into the global economy by a financial liberalization program following the trend
of other Latin-American economies (Carvalho et al., 2014). The program eased
restrictions for foreign participation in the banking sector, established a kind of
universal banking scheme intended to reduce specialization, and implemented financial regulatory measures to promote competition and efficiency in the financial
sector.1 As a result, by 1997 most of state-owned banks were privatized. The share
of public banks in the total assets of the financial system dropped from 43% to
13%, the number of financial institutions increased from 91 in 1990 to 155 in 1997
and the ratio of credit to GDP increased from 30% to 44% (Uribe and Vargas,
2002).
1
Colombian banks are not allowed to offer some financial services that are included in the
standard universal banking approach such as insurance and trust activities.
BANCO DE ESPAÑA
9
DOCUMENTO DE TRABAJO N.º 1537
Evidence has shown that the financial liberalization process in Colombia had
positive consequences by increasing competition and efficiency, lowering intermediation costs and improving loan quality. However, after some years the greater
competition with foreign banks resulted in higher risk levels and a subsequent
deterioration of loans quality, especially among domestic banks (Barajas et al.,
2002). In 1999 the Colombian banking sector was affected by local and external
shocks that triggered the financial conditions in the sector and lead to a profound
financial crisis. The external shock from the Asian financial crisis led to capital
outflows and a rapid depreciation of the exchange rate. At local level, an economic
downturn, and the raise of real interest rates forced to a rapid deterioration of loan
quality that eroded the solvency of the sector. Previous studies reveal that this
rapid deterioration of the financial sector was mainly a consequence of low loan
loss provisions and low capitalization levels (Gomez-Gonzalez, 2009). Between
1998 and 2001 several banking institutions failed, and other were merged. Institutions specialized in mortgage loans were absorbed by large commercial banks.
In consequence, the number of banking institutions felt from 100 in 1998 to 57 in
2001; while the annual rate of credit growth declined from 30% to -6% during the
same period.
Following the financial crisis, Colombian financial authorities strengthened the
regulatory measures intended to enhance adequate provisions for loan losses, and
higher capital and liquidity ratios. These regulatory measures were designed under
the Basel standards with the aim of accounting for the interaction of credit risk
with liquidity and market risk.
Since 2002, risky loans (based on internal loans ratings) were designated as the
target measure to set banks provisions for loan losses, rather than the traditional
non-performing loans (NPL). Thus, loan provisions were settled on an ex-ante
measure of credit risk instead of being computed using an ex-post measure of
credit risk (i.e NPL).2 Market risk was defined as an estimated value by each bank
using the Value at Risk (VaR) of its securities portfolio, which was included as an
additional component in the capital ratio since 2008. Hence, the higher the market
exposure the larger the required capital for the solvency ratio.3 New definitions
of equity capital were also implemented to enhance quality of capital (Tier 1 and
Tier 2). Finally, a short-run liquidity ratio (LR) was required for banks to hedge
from liquidity mismatches.4
2
Provisions vary according to borrowers rating from 1% for type A borrowers up to 20% for
type E borrowers.
3
Capital ratio (CR) should be greater than 9% and is defined as equity capital (CE) over risky
weighted assets (RWA) plus 100/9 of the (VaR). Formally, CR = CE/[RW A + (100/9)(V aR)],
where CR > 9%.
4
LR is the value of liquid assets over short-term liabilities. LR should be positive for maturities
of 7 and 30 days, although it can be negative for 14 days maturities in order to account for the
reserve requirement that banks have to fulfill every two weeks. Previous to LR, regulators used
a ratio of liquid assets over volatile liabilities.
BANCO DE ESPAÑA
10
DOCUMENTO DE TRABAJO N.º 1537
Overall, the above-mentioned regulatory measures have served to influence
banks behavior due to the incorporation of risk-taking. These measures along
with other macroprudential policies implemented in 2006-07 played an important
role to avoid the contagion from the global financial crisis of 2007-08.5 Nevertheless, as we show further, an important decrease in both cost and profit efficiency
was observed during that period, especially for small and foreign banks.
During the period 2002-2012 the Colombian banking sector experienced a growing expansion that has been accompanied by the arrival of foreign banks. The
aggregated value of loans grew 300% and the investments to assets ratio doubled.
Banks increased their competition in the securities market with non-banking institutions (i.e. Brokerage firms) and also boosted their participation in the money
market for short-term liquidity. Several M&A processes were also carried out,
concentrating financial services in a few but large institutions. As a result, risk exposures presented important increases.6 This has required the regulator to closely
monitor credit and market risk and to face the challenges of systemic financial
institutions (see León et al., 2012).
Figure 1 shows the evolution of ratios related to credit, liquidity, capital and
market risk over the period 2002-2012 by distinguishing between small and large
banks and foreign and domestic banks.7 In general, Colombian banks exhibit a
downward trend in credit and market risk along with stable levels of capitalization
and growing liquidity. However, important differences in the level of risk exposure
of banks with different characteristics of size and ownership are observed, which
coincide with the aforementioned regulatory changes including those adopted in
2007-08 to meet Basel II standards. We observe that the ratio of risky loans over
total loans has declined for all banks although large and domestic banks exhibit
higher levels than small and foreign banks. This decreasing trend even during a
period of large credit expansion and high economic growth may be related with the
introduction of the use of risky loans as an indicator for loan losses provisions in
2002. The ratio of liquid assets over total assets has gradually increased over time,
especially for large and foreign banks. Capital ratio seems to be stable for large
5
The government settled limits to banks positions in foreign currency and extended to two
years the period for allowing foreign capital outflows.
6
In May 2013, Colombian Treasury Bill (TES) prices decreased 20% in two weeks as a result
of the uncertainty related to FED’s exit strategy. This led to bank losses of COP 2.32 billion
that represented 4.87% of their equity capital.
7
We define small and large banks as those below and above the median of the total assets
level, respectively.
BANCO DE ESPAÑA
11
DOCUMENTO DE TRABAJO N.º 1537
Liquidity
ratio
Risky loans
ratio
Figure 1: Evolution of risk-exposure measures by type of bank 2002 - 2012
Small vs Large Banks
0.3
0.2
0.2
0.1
0.1
0
2002
1.4
0
2002
1.4
1.2
1.2
Capital
ratio
1
2002
Securities
ratio
Domestic vs Foreign Banks
0.3
2004
2004
2006
2006
2008
2008
2010
2010
2012
2012
1
2002
2004
2006
2008
2010
2012
2004
2006
2008
2010
2012
2004
2006
2008
2010
2012
2006
2008
2010
2012
0.3
0.2
0.1
0
2002
0.4
2004
2006
2008
2010
2012
0.2
0
2002
0.3
0.2
0.1
0
2002
0.4
0.2
2004
2006
Small
2008
Large
2010
2012
0
2002
2004
Domestic
Foreign
banks in Colombia while important increases are observed for small and foreign
banks from 2008. Likewise, small banks reduced more than large institutions
their holdings of securities after the global financial crisis. This may suggest that
small banks were more concerned about the effects of exposures in credit and
securities markets due to the lower probability of being saved given their size, which
made them to highly increase their capital ratios and diminish their market risk
exposures (see Berger and Bowman, 2013, for similar findings in the US banking
sector).
2.1. Efficiency of the Colombian Banking Sector
Early studies of banking efficiency have found evidence of low cost efficiency
in the Colombian banking sector during the 90’s although some improvements
during the first half of 2000s in merged banks (Estrada and Osorio, 2004; Clavijo
et al., 2006). Recent studies have provided evidence on improvements in technical efficiency and productivity in the sector but large heterogeneity among banks.
Sarmiento et al. (2013), using a non-parametric frontier model, found that Colombian banks improved in technical efficiency from 2000 up to the global financial
crisis of 2007-08, when efficiency and productivity decreased. They also found
M&A to have a significant and positive impact on bank efficiency, and high heterogeneity in efficiency irrespective of banks’ size and affiliation. Galán et al.
(2015) estimated input-oriented technical efficiency during the period 2000-2009
BANCO DE ESPAÑA
12
DOCUMENTO DE TRABAJO N.º 1537
using a dynamic Bayesian SFA model. They found out that foreign ownership has
positive and persistent effects on efficiency of Colombian banks, while the effects
of size are positive but rapidly adjusted. They also identified high inefficiency
persistence and important differences between institutions. In particular, merged
banks were found to exhibit low costs of adjustment that allowed them to rapidly
recover the efficiency losses derived from merging processes.
Finally, Moreno and Estrada (2013) studied the role of market power in explaining efficiency gains in Colombian banks during the 2004-2012 period. By
using SFA and non-parametric models, they found a positive relationship between
market power and efficiency, which is explained by product differentiation that allows banks to gain efficiency while not charging excessive credit prices. However,
previous applications have not studied the influence of risk-taking on efficiency of
Colombian banks.
3. Methodology
Frontier efficiency methods have become a very important tool to identify relevant bank inefficiency drivers and to provide useful indicators of performance of
the sector and individual institutions. In particular, SFA, firstly introduced by
Aigner et al. (1977) and Meeusen and van den Broeck (1977), presents the advantages of allowing inferences on the parameters, accounting for idiosyncratic errors
and modeling firm characteristics that affect directly the inefficiency in a single
stage.8 In this context, bank characteristics related to their risk exposures can be
easily and consistently accounted for in cost and profit efficiency estimations.
3.1. Heterogeneity and Risk in Bank Efficiency Measurement
Distinguishing inefficiency from heterogeneity is an important issue in the efficiency frontier literature. Omitting heterogeneity variables has been identified
to lead to biased estimations of inefficiency (see Greene, 2005). In the banking
literature, Bos et al. (2009) identify these effects on efficiency levels and rankings
when observed heterogeneity is omitted. In particular, in the case of risk exposure,
Radić et al. (2012) evaluate a sample of 800 investment banks of G-7 countries
during the period 2001-2007 and find that omitting bank risk-taking from efficiency estimations leads to underestimating profit efficiency. The authors also find
liquidity and capital risk exposures to be the most relevant factors determining
cost and profit inefficiency.
8
In contrast, the main nonparametric method of Data Envelopment Analysis is more flexible
but provides, in general, deterministic measures for inefficiency and does not allow accounting
for inefficiency heterogeneity in a consistent single stage.
BANCO DE ESPAÑA
13
DOCUMENTO DE TRABAJO N.º 1537
Unobserved heterogeneity has also been found to affect estimations from stochastic frontier models.9 In applications to the banking sector, Feng and Zhang (2012)
find that failure to consider unobserved heterogeneity results in misleading efficiency rankings and mismeasured technical efficiencies, productivity growth, and
returns to scale. Goddard et al. (2014) compare different fixed effects, random
effects and random parameters models in an application to Latin American banks
between 1985 and 2010. They find that random parameters models perform better
in distinguishing heterogeneity from inefficiency as well as important differences on
cost efficiency estimations. Williams (2012) applies a random parameters model
in order to test the quiet life hypothesis in Latin American banks. However, the
fact that the author estimates a second stage where cost efficiency is regressed
on a market power index and other bank characteristics may lead to biased and
inconsistent efficiency estimations (see Wang and Schmidt, 2002).
In this context, our proposal is intended to model unobserved heterogeneity
sources related to risk exposures and to account for bank characteristics in a single
stage. Our approach is close to that in Goddard et al. (2014); Williams (2012) in
the use of random parameters in the inefficiency component. However, we propose
to estimate as random the coefficients associated to the observed covariates in the
inefficiency distribution. This allows us to obtain in a single stage bank-specific
estimates of the effects of risk-exposure measures on cost and profit inefficiency.
This specification is more flexible than imposing interactions of observed covariates
with different groups of banks.
3.2. A Stochastic Frontier Model with Random Inefficiency Coefficients
Since we are interested in identifying unobserved heterogeneity related to the
effects of risk on bank inefficiency, we propose a stochastic frontier model where the
coefficients of risk-exposure measures in the inefficiency distribution are modeled
as bank-specific parameters. The proposed specification is the following:
yit = xit β + vit − uit
vit ∼ N (0, σv2 )
uit ∼ Exp(λit )
λit = exp(zit γ + zit∗ γi∗ ),
(1)
where yit represents the output for firm i at time t, xit is a row vector that contains
the input quantities, β is a vector of parameters, vit is an idiosyncratic error
9
Greene (2005) proposes different methods to deal with this kind of heterogeneity under the
frequentist approach. In the Bayesian context, Galán et al. (2014) propose the inclusion of a
random parameter in the inefficiency component that can be modeled along with other observed
covariates and performs well in capturing latent heterogeneity.
BANCO DE ESPAÑA
14
DOCUMENTO DE TRABAJO N.º 1537
assumed to follow a normal distribution, and uit is the inefficiency component. The
inefficiency is assumed to follow an exponential distribution with a firm specific
and time-varying parameter λit , γ is a vector of parameters which are common
to all firms, including the constant, and γi∗ is a vector of firm-specific parameters
intended to capture differences in the effects of covariates across firms on the
inefficiency. Therefore, zit is a vector of heterogeneity variables whose effects
are assumed to be constant across firms, and z∗it contains a set of heterogeneity
variables with bank-specific effects.
This specification with random coefficients in the parameter of the inefficiency
distribution is flexible in the sense that some covariates can be associated with
firm specific coefficients while other heterogeneity variables may be modeled with
fixed coefficients. In particular, the random coefficients are intended to capture
differences in the way risk exposures affect efficiency of different types of banks.
Thus, the model is able to identify not only the effects of risk on inefficiency but
also the type of banks that are more affected by each of the risk-exposure measures.
3.3. Bayesian Inference
The inference of the model is carried out using Bayesian methods. This approach was introduced in stochastic frontier models by van den Broeck et al. (1994)
and allows us to formally incorporate parameter uncertainty and derive posterior
densities of cost and profit efficiency for every individual bank.
We assume proper but relatively dispersed prior distributions throughout. In
particular, the distributions assumed for the parameters in the frontier are: β ∼
N (0, Σβ ) where Σ−1
β is a precision diagonal matrix with priors set to 0.001 for all
coefficients. The variance of the idiosyncratic error term is inverse gamma, which
is equivalent to σv−2 ∼ G(aσv−2 , bσv−2 ) with priors set to 0.01 for the shape and rate
parameters, respectively.
Regarding the inefficiency component, its distribution is assumed to be exponential: uit |γ, γ ∗ , zit , z∗it ∼ Exp(exp(zit γ + z∗it γi∗ )). The prior distribution of
the vector of common parameters is γ ∼ N (0, Σγ ) with priors for the diagonal
precision matrix Σ−1
γ equal to 0.1 for all the coefficients. For the firm-specific
inefficiency heterogeneity coefficients, a hierarchical structure is defined, where
γi∗ ∼ N (γ ∗ , Σγ ∗ ) and γ ∗ is defined in the same way as γ. Sensitivity analysis is performed to the use of an exponential prior distribution for the inefficiency parameters. In this case they are chosen to be centered in a given prior
mean efficiency value r∗ following the procedure in Griffin and Steel (2007), where
exp(γ) ∼ Exp(− ln r∗ ).10 Results show convergence to roughly the same values
after the number of iterations described below.
10 ∗
r is set at 0.65, following other Bayesian SFA studies in banking (see Tabak and Tecles,
2010).
BANCO DE ESPAÑA
15
DOCUMENTO DE TRABAJO N.º 1537
Markov Chain Monte Carlo (MCMC) methods and in particular the Gibbs
Sampling algorithm with data augmentation, as presented by Koop et al. (1995)
for stochastic frontier models, can be used here.11 The MCMC algorithm involves
50,000 iterations where the first 10,000 are discarded and a thinning equal to 4
is used to remove autocorrelations. Therefore, 10,000 iterations are used for the
posterior inference.
We assess the fit and predictive performance of the different models using a
version of the Deviance Information Criterion (DIC) called DIC3 and the Log Predictive Score (LPS) (see Griffin and Steel, 2004; Galán et al., 2014, for applications
of these criteria to Bayesian SFA models). The former is a stable variant of the
within sample measure of fit introduced by Spiegelhalter et al. (2002) commonly
used in Bayesian analysis. Defining the deviance of a model with parameters θ as
D(θ) = −2 log f (y|θ), where y is the data, then DIC = 2D(θ) − D(θ̄). However,
using an estimator of the density f (y|θ) instead of the posterior mean θ̄ is more
stable. This alternative specification presented by Celeux et al. (2006) overcomes
robustness problems when the original DIC is implemented to random effects and
mixture models. The formulation for this criterion is:
DIC3 = −4Eθ [log f (y|θ)|y] + 2 log f (y)
(2)
Regarding LPS, it is a criterion for evaluating the out-of-sample behaviour
of different models. This criterion was first introduced by Good (1952) and is
intended to examine model performance by comparing its predictive distribution
with out-of-sample observations. For this purpose the sample is split into a training
and a prediction set. Our prediction set consists of observations corresponding to
the last two observed years of every firm in the sample, and the training set contains
all the rest. The formula is the following:
1
LP S = −
log f (yi,ti |previous data),
k i=1
k
(3)
where yi,ti represents the observations in the predictive set for the k firms in the
sample and ti represents the penultimate time point with observed data for firm i.
3.4. Translog Cost and Profit Models
We use cost and profit functions for the frontier specification in (1), and we
represent them with translog multi-product functions. The estimated model is:
11
The implementation of our models is carried out using the WinBUGS package (see Griffin
and Steel, 2007, for a general procedure).
BANCO DE ESPAÑA
16
DOCUMENTO DE TRABAJO N.º 1537
R
M M
1
ln cit = β0 + M
β
ln
y
+
δ
ln
p
+
β ln ymit ln ynit
m
m
r
r
it
it
r=1
R m=1
R
M 2 R m=1 n=1 mn
1
+ 2 r=1 s=1 δrs ln prit ln psit + m=1 r=1 ηmr ln ymit ln prit + κ1 t
R
φ
t
ln
y
+
+ 12 κ2 t2 + M
m
m
it
m=1
r=1 ϕr t ln prit + vit + uit
2
vit ∼ N (0, σv )
uit ∼ exp(λit )
J
∗ ∗
λit = exp(γ0 + H
γ
z
+
h
h
it
h=1
j=1 γji zjit−1 ),
(4)
where cit is the total cost or the total profit, y are outputs, p are input prices, and
t is a time trend in order to account for technological change. We also account
for two types of heterogeneity variables affecting cost and profit inefficiency: A
group of bank characteristics modeled in z, which are assumed to have common
effects on all banks, and a group of variables z ∗ , capturing banks’ risk-exposure in
the previous period and allowed to have specific effects on the inefficiency of each
bank. In order to overcome the problem of calculations of logarithms of negative
profits, we follow the rescaling method (Berger and Mester, 1997) which corrects
profit values by a factor Θ equal to the absolute value of the lowest profit plus one.
Linear homogeneity of the cost function is achieved by normalizing total costs and
input prices by a chosen input price. Symmetry of the cross-effects is accomplished
by imposing βmn = βnm , δrs = δsr . In the case of the profit function the sign of
the inefficiency component u is reversed.
From (4) cost efficiency of individual banks in each period is computed as:
CEit = exp(−uit ).
(5)
In the case of profit efficiency, since a constant has been added in order to
allow for profit losses, the efficiency computation is the following (see Berger and
Mester, 1997):
exp(ln π(pit , yit , t; β) − uit ) − Θ
P Eit =
.
(6)
exp(ln π(pit , yit , t; β)) − Θ
Returns to scale (RT S) can be derived from the cost function as the sum of
output elasticities as follows:
RT S =
M
∂ ln C(p, y, t)
∂ ln ym
m=1
−1
,
(7)
where a RT S measure less than 1 indicates that the production technology present
decreasing returns to scale. On the other hand, increasing returns to scale are
observed if the RT S measure is larger than 1, while if it is equal to 1 it indicates
constant returns to scale.
BANCO DE ESPAÑA
17
DOCUMENTO DE TRABAJO N.º 1537
Finally, technical change (T C) assuming constant returns to scale is given by:
M
∂ ln C(p, y, t)
.
(8)
TC = −
∂t
m=1
4. Data
We employ annual data from 31 commercial banks for the period 2002-2012.
This is an unbalanced panel data set from the local central bank (Banco de la
República) and the financial supervisory agency (Superintendencia Financiera de
Colombia). We follow the financial intermediation approach in which banks employ deposits, labor, and physical capital to produce loans, securities investments,
and other financial services.12 We consider as input prices: the price of deposits
(p1 ), which is the ratio of interest expenses divided by total deposits; the price
of labor (p2 ), which is personnel expenses divided by the total number of employees; and the price of physical capital (p3 ), which is calculated as the ratio of
operating expenses (i.e. non-interest reduced by personnel) to total fixed assets.
Thus, these are prices per unit of input. As outputs we consider: loans (y1 ) including consumer, commercial, mortgage, and microcredit; securities (y2 ), which
includes public and private bonds holdings, and other securities investments; and
off-balance-sheet (OBS) activities (y3 ) measured as the ratio of non-interest income over total income. Non-interest income includes securitization, brokerage
services, and management of financial assets for clients, which represent an important source of income for Colombian banks.13 Total costs are considered as the
sum of interest and non-interest costs and total profit as the earned net profit.
We consider two bank-specific characteristics with common effects on the inefficiency of all banks. Those are, size (z1 ), measured as the level of total assets;
and foreign ownership (z2 ), which is a binary variable taking the value of 1 if more
than 50% of bank shares are foreign owned; and 0 otherwise. As aforementioned,
these effects have been found to be relevant inefficiency drivers in previous studies.
As risk-exposure measures with heterogeneous effects on bank-specific inefficiency, we include measures for credit risk, liquidity, capital, and market risk in
accordance with Colombian financial regulation and Basel III standards. Credit
risk (z1∗ ) is measured as risky loans over total loans.14 This measure of ex-ante
12
Hughes and Mester (1993) provide evidence that confirm that deposits should be treated as
inputs (see Sealey and Lindley, 1977, for a discussion on the intermediation approach).
13
In a recent study, Tabak and Tecles (2010) find that omitting OBS as an output over(under-)estimate cost (profit) efficiency results.
14
Risky loans are based on internal loan ratings performed by banks according to the Colombian regulation. This measure of ex-ante credit risk has been used before in the literature to
identify bank risk-taking in the credit market (see Ioannidou and Penas, 2010).
BANCO DE ESPAÑA
18
DOCUMENTO DE TRABAJO N.º 1537
credit risk may avoid biased efficiency estimations that have been identified when
using ex-post credit risk measures such as NPL (see Malikov et al., 2014). Liquidity (z2∗ ) is measured as the ratio of liquid assets over total assets, where liquid
assets include cash holdings, negotiable and available to sell public and private
debt instruments and pledged collateral in repurchase agreement operations. Capitalization (z3∗ ) is measured as the ratio of capital equity over total assets. Finally,
market risk exposure (z4∗ ) is measured as securities investments over total assets.
All risk variables are included lagged one-period in order to account for intertemporal effects on inefficiency and avoid reverse causality.
Table 1 exhibits the summary statistics of the main variables described above,
where all monetary values are expressed in thousands of U.S. dollars at constant
prices from the year 2012.
Table 1: Summary statistics
Variable
Mean
SD
Min
Total loans
3,342,012 4,206,436
11,553
Securities
1,265,349 1,339,794
563
OBS
0.0354
0.0299
0.0266
Price of deposits
0.0248
0.0121
0.0009
Price of labour
36.44
22.30
3.13
Price of capital
1.92
2.66
0.29
Total assets
5,503,680 6,425,746
39,699
Credit risk exposure
0.0988
0.0667
0.0019
Liquidity ratio
0.2296
0.0945
0.0377
Capital ratio
0.1211
0.0757
0.0448
Market risk exposure
0.2381
0.1368
0.0013
Total cost
1,132,776 1,402,621
15,673
Total profit
76,927
377,974 - 784,642
Max
28,267,020
6,461,458
0.0587
0.0923
142.03
17.30
41,786,469
0.3839
0.8226
0.7854
0.7478
7,722,227
2,809,771
Source: Colombian central bank and financial supervisory agency.
5. Results
For comparison purposes, we estimate three different cost (C1 to C3) and
profit (P1 to P3) models from our proposed specification in (4) by including some
restrictions. Models C1 and P1 do not include risk-exposure variables in the
∗
∗
∗
∗
inefficiency, so γ1i
, γ2i
, γ3i
, γ4i
= 0. Models C2 and P2 include the risk covariates
in the inefficiency but restrict them to have a common effect on the inefficiency for
∗
∗
∗
∗
all banks; thus, γ1i
, γ2i
, γ3i
, γ4i
= γ1∗ , γ2∗ , γ3∗ , γ4∗ . Finally, our proposed specification
to model random inefficiency coefficients is estimated in models C3 and P3.
BANCO DE ESPAÑA
19
DOCUMENTO DE TRABAJO N.º 1537
Given that our interest is to analyze the effects of size, ownership and riskexposure on efficiency, we present the estimation results only for the parameters in
the inefficiency distribution. Tables 2 and 3, present the posterior mean and probability intervals for the parameters in the cost and profit inefficiency component,
respectively. Results for the frontier parameters are presented in the Appendix in
Tables A.1 and A.2.15
Table 2: Posterior mean and 95% probability intervals of parameters in the inefficiency distribution of cost models
Model C1
No risk covariates
Mean
95% PI
γ0
γ1 size
γ2 f oreign
γ1∗ credit
γ2∗ liquidity
γ3∗ capital
γ4∗ market
Mean efficiency
s.d. efficiency
DIC3
LP S
1.0350
-0.1981
-0.8144
[0.0142, 0.0591]
[-0.2913, -0.0820]
[-1.9580, -0.0943]
0.8934
0.0653
2982.76
-9.62
Model C2
Common risk coefficients
Mean
95% PI
0.9046
-0.1823
-0.6358
0.3058
0.2962
-1.0823
0.0341
[0.4875, 1.4407]
[-0.3041, -0.0762]
[-1.1525, -0.1482]
[0.0845, 0.5280]
[-0.0598, 0.6511]
[-1.5908, -0.5961]
[-1.0370, 1.0452]
0.8923
0.1466
2416.44
-61.57
Model C3
Random risk coefficients
Mean
95% PI
0.8925
-0.1595
-0.2198
0.2863
0.3511
-1.9502
-0.0054
[0.4532, 1.3128]
[-0.3070, -0.0109]
[-0.4206, -0.0693]
[0.0932, 0.5816]
[-0.0718, 0.7160]
[-2.8485, -1.0214]
[-1.0517, 1.0326]
0.7102
0.2251
2005.79
-91.79
Note: Values for γ1∗ to γ4∗ in Model C3 correspond to the average posterior distribution of individual coefficients.
5.1. Model Comparison
Model comparison indicators lead to similar conclusions in both the cost and
profit models.16 That is, models including measures of risk exposure improve from
models omitting these variables (C1 and P1). This suggests that risk-taking is
an important determinant of bank inefficiency. From the models considering risk
exposures, those including random coefficients for the risk covariates in the inefficiency distribution (C3 and P3) exhibit the best fit and predictive performance.
These results suggest not only that measures of risk exposure are important inefficiency drivers but also that risk has different effects on cost and profit inefficiency of
15
From the frontier parameter estimates, it is observed that loans, investments, and OBS
positively affect cost and input prices in all models. In the case of profits, the relationship is also
positive for loans and investments but negative, although not significant, for OBS. This result
for OBS was also found by Tabak and Tecles (2010) in an application to the Indian banking
sector. However, they found loans and investments to be not significant when OBS is included
in both cost and profit models. Regarding input prices, the coefficients are not relevant in the
profit models.
16
Lower values for DIC3 and LP S indicate better fit and predictive performance.
BANCO DE ESPAÑA
20
DOCUMENTO DE TRABAJO N.º 1537
Table 3: Posterior mean and 95% probability intervals of the inefficiency parameter distributions
in profit models
Model P1
No risk covariates
Mean
95% PI
γ0
γ1 size
γ2 f oreign
γ1∗ credit
γ2∗ liquidity
γ3∗ capital
γ4∗ market
Mean efficiency
s.d. efficiency
DIC3
LP S
-0.9668
0.0556
1.0120
[-2.4390, 0.3062]
[0.0226, 0.1491]
[0.6873, 1.3594]
0.5150
0.1638
3168.01
-180.01
Model P2
Common risk coefficients
Mean
95% PI
-1.2045
0.0779
1.0347
-2.0264
0.1925
-1.4811
-0.8884
[-2.7145, 0.3825]
[0.0321, 0.1316]
[0.6906, 1.3790]
[-2.8005, -1.3274]
[-0.6021, 1.0655]
[-2.0741, -0.6972]
[-1.5174, -0.2640]
0.6205
0.2281
2458.10
-302.42
Model P3
Random risk coefficients
Mean
95% PI
-1.4024
0.1277
1.0200
-1.5812
0.5437
-1.4367
-0.9531
[-2.9651, 0.2635]
[0.0054, 0.1927]
[0.4760, 1.5287]
[-2.5727, -0.6431]
[-0.3528, 1.3919]
[-2.1599, -0.5821]
[-1.6614, -0.2650]
0.6714
0.3281
2360.85
-405.94
Note: Values for γ1∗ to γ4∗ in Model P3 correspond to the average posterior distribution of individual coefficients.
banks with different characteristics. This has important implications for efficiency
estimations. In tables 2 and 3, we observe that posterior mean cost and profit
efficiency are over- underestimated, respectively, and that their dispersion is lower
when risk-exposure measures are not modeled as bank-specific in the inefficiency
distribution.
Figure 2 exhibits also differences in the predictive efficiency distributions of cost
and profit models. We observe that both location and dispersion of the distributions are affected (see Koetter, 2008, for similar results)). In particular, predictive
distributions from models including risk in the inefficiency are more symmetric
and those derived from models with random coefficients present less dispersion.
Overall, these results evidence the importance of accounting for risk-taking and
its associated heterogeneity among banks when estimating bank inefficiency (see
Hughes et al., 2001; Koetter, 2008; Pessarossi and Weill, 2015; Radić et al., 2012;
Malikov et al., 2014, for previous evidence).
5.2. Inefficiency Determinants
We observe that size and foreign ownership are important inefficiency drivers
in all the models. Their effects are negative on cost inefficiency and positive on
profit inefficiency. Previous studies have found similar effects. Chen and Liao
(2011) found that foreign banks perform better than local banks because they
may better deal with risk exposures given cheaper access to funding sources and
more diversification. Fries and Taci (2005) found similar results for banks with
a majority of foreign ownership in emerging economies. Regarding size, previous
studies have found that large institutions tend to exhibit greater efficiency associated with higher scale economies (Bos and Kool, 2006; Wheelock and Wilson,
BANCO DE ESPAÑA
21
DOCUMENTO DE TRABAJO N.º 1537
Figure 2: Predictive distributions of efficiency under cost and profit models
Cost models
25
20
15
C1
C2
C3
10
5
0
0
0.1
0.2
0.3
0.4
0.6
0.7
0.8
0.9
1.0
0.6
0.7
0.8
0.9
1.0
Profit models
25
20
0.5
P1
P2
P3
15
10
5
0
0
0.1
0.2
0.3
0.4
0.5
This figure depicts differences in the predictive efficiency distributions of cost and profit models. Models with no
risk-taking covariates (C1 and P1), models with common risk coefficients (C2 and P2) and models with random
risk coefficients (C3 and P3).
2012; Hughes and Mester, 2013). In previous applications to Colombian banks,
both foreign and large banks have also been found to be more cost efficient than
local and small banks (Moreno and Estrada, 2013; Sarmiento et al., 2013; Galán
et al., 2015).
This relative advantage of large over small banks has been recently reported in
the literature as evidence of the too-big-to-fail dilemma where larger banks take
advantage of their size to obtain funds at lower cost and take on more risk (Santos,
2014). Bertay et al. (2013) analyzed a large sample of banks for 90 countries during
the period from 1992-2011 and found that bank interest costs tend to decline with
systemic size.
Size and foreign ownership are also key characteristics determining the way
credit and market risk, and liquidity and capitalization levels affect cost and profit
efficiency. This is identified through the random coefficient models. We analyze
these effects by type of banks (i.e. small vs. large and domestic vs. foreign).
Figures 3 and 4 present 95% probability intervals of average posterior random
coefficients by type of bank in the cost and profit models, respectively. We observe
two main results when bank-specific coefficients are estimated. First, some groups
of banks are more affected than others taking the same risk exposures. Second,
the effects of risk exposures become relevant as inefficiency drivers for some types
of banks.
BANCO DE ESPAÑA
22
DOCUMENTO DE TRABAJO N.º 1537
5.2.1. Credit Risk
Credit risk is identified as a key determinant of both cost and profit inefficiency
though with opposite effects. While credit risk is found to have positive effects on
cost inefficiency (i.e. negative effects on cost efficiency), it affects negatively profit
inefficiency (i.e. positive effects on profit efficiency). These results are identified in
both the fixed and the random coefficients models and may suggest that assuming
higher credit risk exposures implies expending more resources on monitoring and
administering problem loans. Berger and DeYoung (1997) also found evidence
on this negative effect of problem loans on cost efficiency in U.S. banks and argue
that extra costs may be represented by additional monitoring, negotiating possible
workout arrangements, disposing collateral for possible defaults, defending bank’s
safety to the market and supervisor, and additional precautions to reserve quality
of other loans. On the other hand, in term of profits banks earn extra profits from
riskier loans and may have incentives to engage in higher credit risk.
By type of banks, we identify important differences in the way credit risk affect
efficiency. Large and domestic banks are found to be less affected in cost efficiency
by assuming the same level of credit risk. That is, it is less costly for large and
domestic banks to manage problem loans. A possible explanation could be related
to the fact that local banks have better information about borrowers which implies
that these banks may incur in lower monitoring costs. As to large banks, they may
benefit from scale economies that allows them to incur proportionally in lower costs
at the same credit risk levels. Regarding profit efficiency, large and foreign banks
benefit more from assuming similar levels of credit risk. These types of banks
may take advantage of their recognition in order to charge higher interest rates
for loans of similar quality or are exploiting market power benefits (see Boyd and
De Nicoló, 2005; Wagner, 2010).
5.2.2. Liquidity
Although results from our models with common coefficients suggest that liquidity does not have relevant effects on efficiency of Colombian banks, the random
coefficients model identifies an important positive effect of liquidity on cost inefficiency (i.e. negative effect on cost efficiency) of domestic banks. This suggests
that holding the same proportion of liquid assets is more costly for local banks.
This could be explained by the fact that foreign banks may have greater access to
interbank markets and to cheaper sources of funding (Chen and Liao, 2011). In
the case of profit efficiency, no differences are found in the way liquidity affects
efficiency of banks with different characteristics of size and ownership.
BANCO DE ESPAÑA
23
DOCUMENTO DE TRABAJO N.º 1537
Figure 3: Probability intervals of risk-exposure coefficients by type of bank in cost efficiency
model C3
Credit risk
0
0.2
0.4
0.6
Liquidity
0.8
Small
Small
Large
Large
Domestic
Domestic
Foreign
Foreign
1
−0.5
−0.25
0
Capital
−3
−2.5
−2
−1.5
0.25
0.5
0.75
1
Market risk
−1
Small
Small
Large
Large
Domestic
Domestic
Foreign
Foreign
−0.5
−1.5
−1
−0.5
0
0.5
1
1.5
Note: 95% probability density intervals of average posterior distributions of the random inefficiency coefficients.
5.2.3. Capital
We identify that higher capitalization levels lead to higher cost and profit
efficiency. Reasons behind these results may be derived from the agency problems
between shareholder and managers. Shareholders of highly capitalized banks have
more incentives to control better costs and capital allocation than shareholders of
thinly capitalized banks. This behavior incentivizes managers to put in practice
cost reducing strategies that lead to higher efficiency. Previous studies have also
found evidence showing that highly capitalized banks tend to be more efficient
than thinly capitalized banks (see Kwan and Eisenbeis, 1997; Fiordelisi et al.,
2011; Radić et al., 2012). Berger and DeYoung (1997) also suggest an indirect
effect through credit risk. That is, highly capitalized banks have less moral hazard
incentives to take on higher risk, and therefore they will incur in less costs.
Regarding differences in the effect of capital on efficiency between banks with
different sizes and ownerships, our results may suggest that small and domestic
banks benefit more in terms of cost efficiency. However, it is worth to notice that
the probability that these estimates are lower than those of large and foreign banks
are less than 95%. On this regard, Berger and Bowman (2013) have found that
small banks benefit more than large banks from increases in capital specially during
the financial crisis. Also, Pessarossi and Weill (2015) have found that domestic
banks in China benefit from having higher capital while the effect for foreign banks
BANCO DE ESPAÑA
24
DOCUMENTO DE TRABAJO N.º 1537
is not significant. They argue that domestic banks in China have more government
guarantees in case of financial distress. This would increase agency costs between
shareholders and debtholders, which would become more important than agency
costs between shareholders and managers. In terms of profit efficiency, no relevant
differences are found between banks with different characteristics.
Figure 4: Probability intervals of risk-exposure coefficients by groups of banks in profit efficiency
model P3
Credit risk
−3
−2.5
−2
−1.5
Liquidity
−1
−0.5
Small
Small
Large
Large
Domestic
Domestic
Foreign
Foreign
0
−0.5
0
Capital
−2.5
−2
−1.5
0.5
1
1.5
Market risk
−1
−0.5
Small
Small
Large
Large
Domestic
Domestic
Foreign
Foreign
0
−2
−1.5
−1
−0.5
0
Note: 95% probability density intervals of average posterior distributions of the random inefficiency coefficients.
5.2.4. Market Risk
As to market risk, we find no evidence of important effects on cost inefficiency
of Colombian banks. This result holds when heterogeneous effects are accounted
for in the random coefficients models, suggesting that market risk is not a cost
efficiency determinant for any type of bank. Nevertheless, market risk have important negative effects on banks profit inefficiency (i.e. positive effects on profit
efficiency). Moreover, the random coefficients model shows strong evidence supporting that these effects are more important for large and foreign banks, which
would have greater incentives to engage in more market risk. Large and foreign
banks may benefit from having more diversified portfolios and access to cheaper
funding sources that allow them to get higher returns on their investments (as
reported by Chen and Liao (2011)). Also, large banks are the primary dealers of
the Colombian public debt market. This condition allows them to obtain higher
profits by selling debt bills to small banks, who use them as collateral to obtain
liquidity from the central bank and from the secured money market.
BANCO DE ESPAÑA
25
DOCUMENTO DE TRABAJO N.º 1537
5.3. Efficiency, Technical Change and Returns to Scale
The evolution of cost and profit efficiency over time is presented in Figure
5 by groups of banks. We observe that large and foreign banks exhibit higher
cost efficiency levels than small and local banks. A possible explanation for the
differences between banks with different sizes may be related to the fact that large
banks might be considered by creditors as too-big-to-fail, which allows them to
have access to cheaper funding sources. Small banks have been more volatile in
both cost and profit efficiency over time, specially after the global financial crisis,
while large banks have been more stable and present higher cost efficiency over the
whole period. This may suggest that large banks are less sensitive to environmental
conditions, possibly due to more stable funding sources. In the case of small banks
the result might be the opposite because creditors and depositors may ask for
higher returns from those banks as a way to exert market discipline (see evidence
in Wheelock and Wilson, 2012; Bertay et al., 2013; Hughes and Mester, 2013).
Regarding ownership, although foreign banks present higher cost efficiency than
local banks, in terms of profit efficiency they exhibit lower scores and much more
volatility over the whole period. The highest difference is observed in 2008 coinciding with the global financial crisis. This suggests that foreign institutions were
more affected due to their operations and investments in international markets.
Nevertheless, in the last few years, foreign banks have improved and exhibited an
increasing trend in profit efficiency.
Finally, we compute technical change and returns to scale from Model C3 and
report the results in Table 4 by groups of banks with similar characteristics of size
and ownership. In general, we observe technical progress for all types of banks
but specially for large and domestic institutions. This can be a consequence of the
reorganization processes that these types of institutions have carried out during
the period including several M&A. Regarding returns to scale, decreasing returns
are observed in the Colombian banking sector, which may suggest low margins for
more M&A processes. However, some important differences are found when the
analysis is performed by groups of banks. We find that while large and domestic
institutions operate at decreasing returns to scale, small and foreign banks exhibit
increasing returns to scale. These results coincide with those reported by Galán
et al. (2015), who suggest that M&A processes carried out mainly by domestic and
large institutions may lead them to be oversized, while small and foreign banks
may still present some potential scale gains. Furthermore, the fact that large
banks exhibit decreasing returns to scale may confirm that their efficiency gains
obey to external sources such as lower funding costs (i.e. deposits, subordinated
debt or interbank loans) as a result of implicit government guarantees. On this
BANCO DE ESPAÑA
26
DOCUMENTO DE TRABAJO N.º 1537
Figure 5: Evolution of mean posterior cost and profit efficiency by groups of banks in random
coefficient models
Cost efficiency
Small vs Large Banks
Domestic vs Foreign Banks
1
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
2002
2004
2006
2008
2010
2012
Profit efficiency
1
0
2002
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
2002
2004
2006
2008
Small
2010
2012
0
2002
2004
2006
2008
2010
2012
2004
2006
2008
2010
2012
Domestic
Large
Foreign
regard, Davies and Tracey (2014) evaluated a panel of the largest international
commercial banks over the period from 2001 to 2010 and found that large banks
benefit from implicit subsidies and that suppressing them makes scale economies
disappear. Their results imply that estimated scale economies for large banks are
affected by too-big-to-fail considerations.
Table 4: TC and
Bank type
Small
Large
Domestic
Foreign
All banks
RTS by type of bank
TC
RTS
0.0332 1.0473
0.0522 0.9216
0.0474 0.9211
0.0425 1.0413
0.0456 0.9618
6. Concluding remarks
Risk-taking is an inherent condition of the banking business. However, traditional studies on bank efficiency had assumed that risk is incorporated into bank
output without explicitly modeling its role in explaining inefficiency. Recent studies show that failure to account for risk-taking may lead to biased estimations
BANCO DE ESPAÑA
27
DOCUMENTO DE TRABAJO N.º 1537
of bank efficiency and misleading estimates of scale economies and cost elasticities. Likewise, the literature has focused mainly on credit risk and capitalization,
omitting other important risks faced by banks.
We present a stochastic frontier model with random inefficiency coefficients,
which is able to capture unobserved heterogeneity related to credit, liquidity, capital, and market risk exposures. The model is found to accurately distinguish bank
heterogeneity from inefficiency and provides the first empirical evidence on the role
of bank risk-taking in the inefficiency of the Colombian banking industry. In line
with previous evidence, our findings remark the importance of accounting for size,
affiliation and risk exposure in the estimation of bank efficiency (Bos et al., 2009;
Radić et al., 2012; Goddard et al., 2014; Pessarossi and Weill, 2015). Cost and
profit efficiency are found to be over- underestimated when risk measures are not
accurately modeled into the profit and cost function(Hughes et al., 2001; Koetter,
2008; Malikov et al., 2014). We also find that size and foreign ownership are not
only important determinants of efficiency but also key characteristics determining the way credit and market risk, and liquidity and capitalization levels affect
cost and profit efficiency. The main channels supporting these differences among
banks with different characteristics are related to monitoring costs, diversification,
information asymmetries, agency costs, risk-taking incentives, among others.
We find that higher credit risk exposures lead to lower cost efficiency that can
be associated with greater expenditures on monitoring and administering problem
loans. However, our findings suggest that these costs are lower for large and
domestic banks. Large banks may benefit from scale economies that allows them to
incur proportionally in lower costs at the same credit risk levels, while local banks
may incur in lower monitoring costs given that they have better information about
borrowers. We also find credit risk to be associated with higher profit efficiency
and that large and foreign banks benefit more from assuming similar levels of credit
risk.
We provide evidence to support the hypothesis that capital requirements and
buffers may contribute to enhance banking efficiency (Chortareas et al., 2012;
Barth et al., 2013; Pessarossi and Weill, 2015). We identify that higher capitalization levels lead to higher efficiency in both costs and profits, specially for small
and domestic banks. This can be related to agency problems between shareholders
and managers. Shareholders of highly capitalized banks have more incentives to
control better costs and capital allocation, while managers of these institutions
have less moral hazard incentives to take on higher credit risk. However, we find
that those marginal benefits from capitalization are lower for small banks after
2008 coinciding with the global financial crisis and the regulatory changes on capital ratios and credit risk implemented by the Colombian regulator. This finding
may imply that the positive effect of capitalization on the incentives of both share-
BANCO DE ESPAÑA
28
DOCUMENTO DE TRABAJO N.º 1537
holders and managers to be more efficient can be limited when the cost of raising
capital increases.
Our results also identify positive effects of market risk on profit efficiency.
In particular, large and foreign institutions are found to have greater incentives
to engage in more market risk (Radić et al., 2012). These types of banks may
benefit of having more diversified portfolios and access to cheaper funding sources
that allow them to obtain higher returns on their investments. Large banks also
benefit from being the primary dealers of the Colombian public debt market which
enhances their efficiency gains from the trading activity in this market.
Finally, large banks are found to present higher efficiency than small institutions and to be less affected by the financial crisis (Berger and Bowman, 2013).
Moreover, the fact that large banks face lower costs and present incentives to take
on more risk in credit and securities constitutes a signal for regulators to closely
monitor the behavior of these type of banks and their riskiness. Decreasing returns to scale exhibited by large banks may also suggest that their cost and profit
efficiency gains obey to external sources such as lower funding costs (i.e. deposits,
subordinated debt or interbank loans) as a result of implicit government guarantees (Davies and Tracey, 2014). Thus, regulators should also consider alternative
measures to limit risk-taking incentives associated with the fact that large banks
may benefit of being considered as too-big-to-fail.
Overall, bank efficiency estimations should account for the influence of risktaking along with other bank-characteristics of size and ownership. We study
these interactions by employing random coefficients models which provide a suitable approach to deal with this type of heterogeneity in banking production models.
Banks cost and profit efficiency measures that account for risk-taking may constitute a useful indicator for financial stability considerations given that banks with
lower efficiency have been found to be more prone to future bank fails and tend
to engage on more risk (Berger and DeYoung, 1997; Podpiera and Weill, 2008).
Regulators should be aware not only of the consequences of macroprudential regulation on bank performance, but also of the different effects that polices intended
to discourage risk exposure have on banks with different characteristics.
Acknowledgements
We acknowledge Harry Huizinga and Wolf Wagner for their valuable comments and suggestions. We also thank Subal Kumbhakar, Mike Tsionas, Alberto Ortiz, Óscar Carvallo, Pamela
Cardozo, Sandra Benitez, Esteban Gomez, Daniel Osorio, Rocio Betancourt, Orlando Chipatecua and the participants of the Central Bank Research Joint Seminar on Monetary Policy and
Financial Stability (CEMLA, 2014) and the 51th Eastern Finance Association meeting (EFA,
2015) for their useful comments.
BANCO DE ESPAÑA
29
DOCUMENTO DE TRABAJO N.º 1537
Agoraki, M., M. Delis, and F. Pasiouras (2011). Regulations, competition and bank risk-taking
in transition countries. Journal of Financial Stability 7, 38–48.
Aigner, D., C. Lovell, and P. Schmidt (1977). Formulation and estimation of stochastic frontier
production function models. Journal of Econometrics 6, 21–37.
Altunbas, Y., S. Carbo, E. Gardener, and P. Molyneux (2007). Examining the relationships
between capital, risk and efficiency in European banking. European Financial Management 13
(10), 49–70.
Anginer, D., A. Demirguc-Kunt, H. Huizinga, and M. Kebin (2013). How does corporate governance affect bank capitalization strategies? Policy Research Working Paper 6636, The World
Bank.
Barajas, A., R. Steiner, and A. Salazar (2002). The impact of liberalization and foreign direct
investment in Colombia’s financial sector. Journal of Development Economics 23, 157–196.
Barth, J., C. Lin, Y. Ma, J. Seade, and F. Song (2013). Do bank regulation, supervision and
monitoring enhance or impede bank efficiency? Journal of Banking and Finance 37, 2879–
2892.
Berger, A. and C. Bowman (2013). How does capital affect bank performance during financial
crises? Journal of Financial Economics 109, 146–176.
Berger, A. and R. DeYoung (1997). Problem loans and cost efficiency in commercial banks.
Journal of Banking and Finance 21(6), 849–870.
Berger, A. and L. Mester (1997). Journal of banking and finance. Inside the Black Box: What
Explains Differences in the Efficiencies of Financial Institutions? 21, 895–947.
Bertay, A., A. Demirgüç-Kunt, and H. Huizinga (2013). Do we need big banks? Evidence on
performance, strategy and market discipline. Journal of Financial Intermediation 22, 532–558.
BIS (2010). Basel III: International framework for liquidity risk measurement, standards and
monitoring. Basel Committee on Banking Supervision December, Bank for International
Settlements.
BIS (2011). Basel III: A global regulatory framework for more resilient banks and banking
systems. Basel Committee on Banking Supervision June, Bank for International Settlements.
Bos, J., M. Koetter, J. Kolari, and C. Kool (2009). Effects of heterogeneity on bank efficiency
scores. European Journal of Operational Research 195, 251–261.
Bos, J. and C. Kool (2006). Bank efficiency: The role of bank strategy and local market conditions. Journal of Banking and Finance 30, 1953–1974.
Boyd, J. and G. De Nicoló (2005). The theory of bank risk taking and competition revisited.
Journal of Finance 60, 1329–1343.
Brunnermeier, M. and L. Pedersen (2009). Market liquidity and funding liquidity. Review of
Financial Studies 22 (6), 2201–2238.
BANCO DE ESPAÑA
30
DOCUMENTO DE TRABAJO N.º 1537
Carvalho, F., L. Paula, and J. Williams (2014). Banking in latin america. In A. Berger,
P. Molyneux, and J. Wilson (Eds.), The Oxford Handbook of Banking, Second Edition. Oxford
University Press, Inc.
Celeux, G., F. Forbes, C. Robert, and D. Titterington (2006). Deviance information criteria for
missing data models. Bayesian Analysis 4, 651–674.
Chen, S. and C. Liao (2011). Are foreign banks more profitable than domestic banks? Homeand host-country effects of banking market structure, governance, and supervision. Journal
of Banking and Finance 35, 819–839.
Chortareas, G., C. Girardone, and A. Ventouri (2012). Bank supervision, regulation, and efficiency: Evidence from the European Union. Journal of Financial Stability 8, 292–302.
Clavijo, S., C. Rojas, C. Salamanca, G. Montoya, and C. Rizo (2006). Mergers and acquisitions
in the Colombian financial sector: Impact on efficiency (1990-2005). Icfai Journal of Mergers
and Acquisitions 3, 38–61.
Davies, R. and B. Tracey (2014). Too big to be efficient? the impact of implicit subsidies on
estimates of scale economies for banks. Journal of Money Credit and Banking 46, 219–253.
Estrada, D. and P. Osorio (2004). Efectos del capital financiero en la eficiencia del sistema
bancario colombiano. Ensayos Sobre Polı́tica Económica 47, 162–201.
Feng, G. and X. Zhang (2012). Productivity and efficiency at large and community banks in the
US: A Bayesian true random effects stochastic distance frontier analysis. Journal of Banking
and Finance 36, 1883–1895.
Fiordelisi, F., D. Marques-Ibanez, and P. Molyneux (2011). Efficiency and risk in European
banking. Journal of Banking and Finance 35, 1315–1326.
Fries, S. and A. Taci (2005). Cost efficiency of banks in transition: Evidence from 289 banks in
15 post-communist countries. Journal of Banking and Finance 29, 55–81.
Galán, J., H. Veiga, and M. Wiper (2014). Bayesian estimation of inefficiency heterogeneity in
stochastic frontier models. Journal of Productivity Analysis 42, 85–101.
Galán, J., H. Veiga, and M. Wiper (2015). Dynamic effects in inefficiency: Evidence from the
Colombian banking sector. European Journal of Operational Research 240, 562–571.
Goddard, J., P. Molyneux, and J. Williams (2014). Dealing with cross-firm heterogeneity in bank
efficiency estimates: Some evidence from Latin America. Journal of Banking and Finance 40,
130–142.
Gomez-Gonzalez, J. a. N. K. (2009). Bank failure: Evidence from the colombian financial crisis.
The International Journal of Business and Finance Research 3, 15–31.
Good, I. (1952). Rational decisions. Journal of the Royal Statistical Society B 14, 107–114.
Greene, W. (2005). Reconsidering heterogeneity in panel data estimators of the stochastic frontier
model. Journal of Econometrics 126, 269–303.
BANCO DE ESPAÑA
31
DOCUMENTO DE TRABAJO N.º 1537
Griffin, J. and M. Steel (2004). Semiparametric Bayesian inference for stochastic frontier models.
Journal of Econometrics 123, 121–152.
Griffin, J. and M. Steel (2007). Bayesian stochastic frontier analysis using WinBUGS. Journal
of Productivity Analysis 27, 163–176.
Hughes, J. and L. Mester (1993). A quality and risk-adjusted cost function for banks: Evidence
on the ‘too-big-to-fail” doctrine. Journal of Productivity Analysis 4 (3), 293–315.
Hughes, J. and L. Mester (2013). Who said large banks don’t experience scale economies?
Evidence from a risk-return-driven cost function. Journal of Financial Intermediation 22 (4),
559–585.
Hughes, J., L. Mester, and C. Moon (2001). Are scale economies in banking elusive or illusive?:
Evidence obtained by incorporating capital structure and risk-taking into models of bank
production. Journal of Banking and Finance 25 (12), 2169–2208.
Ioannidou, M. and M. Penas (2010). Deposit insurance and bank risk-taking: Evidence from
internal loan ratings. Journal of Financial Intermediation 19, 95–115.
Koetter, M. (2008). The stability of efficiency rankings when risk-preferences and objectives are
different. The European Journal of Finance 14 (2), 115–135.
Koop, G., M. Steel, and J. Osiewalski (1995). Posterior analysis of stochastic frontier models
using Gibbs sampling. Computational Statistics 10, 353–373.
Kwan, S. and R. Eisenbeis (1997). Bank risk, capitalization and operating efficiency. Journal of
Financial Services Research 12 (2), 117–131.
Laeven, L. and R. Levine (2009). Bank governance, regulation and risk taking. Journal of
Financial Economics 93, 259–275.
León, C., C. Machado, F. Cepeda, and M. Sarmiento (2012). Systemic risk in large value payment
systems in colombia: a network topology and payments simulation approach. In M. Hellqvist
and T. Laine (Eds.), Diagnostics for the financial markets - computational studies of payment
system, Simulator Seminar Proceedings, 2009-2011. Bank of Finland.
Lepetit, L., E. Nys, P. Rous, and A. Tarazi (2008). Bank income structure and risk: an empirical
analysis of European banks. Journal of Banking and Finance 32 (8), 1452–1467.
Malikov, E., D. Restrepo-Tobón, and S. Kumbhakar (2014). Estimation of banking technology
under credit uncertainty. Empirical Economics DOI:10.1007/s00181-014-0849-z, in press.
Meeusen, W. and J. van den Broeck (1977). Efficiency estimation from Cobb-Douglas production
functions with composed errors. International Economic Review 8, 435–444.
Moreno, D. and D. Estrada (2013). Colombian bank efficiency and the role of market structure.
Temas de Estabilidad Financiera, No. 76, Banco de la República.
Pessarossi, P. and L. Weill (2015). Do capital requirements affect cost efficiency? evidence from
china. Journal of Financial Stability 19, 119–127.
BANCO DE ESPAÑA
32
DOCUMENTO DE TRABAJO N.º 1537
Podpiera, J. and L. Weill (2008). Bad luck or bad management? Emerging banking market
experience. Journal of Financial Stability 4 (2), 135–148.
Radić, N., F. Fiordelisi, and C. Girardone (2012). Efficiency and risk-taking in pre-crisis investment banks. Journal of Financial Services Research 41(1), 81–101.
Rajan, R. (2006). Has finance made the world riskier? European Financial Management 12 (4),
499–533.
Santos, J. (2014). Evidence from the bond market on banks’ “too big to fail” subsidy. Economic
Policy Review 20(2), 29–39.
Sarmiento, M., A. Cepeda, H. Mutis, and J. Perez (2013). Nueva evidencia sobre la eficiencia
de la banca colombiana: Una medición con modelos de frontera no-paramétricos. Archivos de
Economı́a 392, Departamento Nacional de Planeación.
Sealey, C. and J. Lindley (1977). Inputs, outputs and a theory of production and cost at
depository financial institutions. Journal of Finance 32, 1252–1266.
Shleifer, A. and R. Vishny (2010). Unstable banking. Journal of Financial Economics 97 (3),
306–318.
Spiegelhalter, D., N. Best, B. Carlin, and A. van der Linde (2002). Bayesian measures of model
complexity and fit. Journal of the Royal Statistical Society 64 (4), 583–616.
Tabak, B., G. Craveiro, and D. Cajueiro (2011). Bank efficiency and default in Brazil: Causality
tests. Working Paper Series, No. 235, Central Bank of Brazil.
Tabak, B. and P. Tecles (2010). Estimating a Bayesian stochastic frontier for the Indian banking
system. International Journal of Production Economics 125, 96–110.
Uribe, J. and H. Vargas (2002). Financial reform, crisis and consolidation in Colombia. Borradores de Economı́a, No. 204, Banco de la República.
van den Broeck, J., G. Koop, J. Osiewalski, and M. Steel (1994). Stochastic frontier models: A
Bayesian perspective. Journal of Econometrics 61, 273–303.
Wagner, W. (2010). Loan market competition and bank risk-taking. Journal of Financial Services
Research 37, 71–81.
Wang, H. and P. Schmidt (2002). One step and two step estimation of the effects of exogenous
variables on technical efficiency levels. Journal of Productivity Analysis 18, 129–144.
Wheelock, D. C. and P. W. Wilson (2012). Do large banks have lower costs? New estimates of
returns to scale for U.S. banks. Journal of Money, Credit and Banking 44(1), 171–199.
Williams, J. (2004). Determining management behavior in European banking. Journal of Banking
and Finance 28(10), 2427–2460.
Williams, J. (2012). Efficiency and market power in Latin American banking. Journal of Financial Stability 8, 263–276.
BANCO DE ESPAÑA
33
DOCUMENTO DE TRABAJO N.º 1537
Appendix
Table A.1: Posterior mean and 95% probability intervals of frontier parameter distributions in
cost models
β0
β1
β2
β3
β11
β12
β13
β22
β23
β33
δ1
δ2
δ11
δ12
δ22
η11
η12
η21
η22
η31
η32
κ1
κ2
φ1
φ2
φ3
ϕ1
ϕ2
BANCO DE ESPAÑA
34
Model C1
No risk covariates
Mean
95% PI
Model C2
Common risk coefficients
Mean
95% PI
Model C3
Random risk coefficients
Mean
95% PI
9.789
3.025
3.391
-0.199
-0.457
0.267
0.017
0.010
-0.003
0.001
-1.678
0.820
0.179
-0.001
-0.077
0.170
0.123
0.005
-0.124
-0.009
-0.006
-0.336
0.007
0.066
-0.023
-0.003
0.049
-0.054
3.698
4.031
4.586
-0.261
-0.511
0.307
0.015
0.029
-0.001
0.001
-4.381
2.272
-0.029
0.095
-0.150
0.319
0.036
-0.044
-0.090
-0.016
0.004
-0.543
-0.005
0.083
-0.022
-0.002
0.044
-0.052
7.205
2.914
3.538
-0.212
-0.399
0.231
0.013
0.048
0.001
-0.001
-3.307
1.645
-0.043
0.083
-0.204
0.235
0.111
-0.030
-0.114
-0.009
-0.003
-0.511
-0.001
0.078
-0.030
-0.002
0.024
-0.044
[-3.249,24.840]
[0.6511,5.195]
[1.609,5.336]
[-0.391,0.013]
[-0.713,-0.201]
[0.0725,0.459]
[0.0024,0.033]
[-0.108,0.146]
[-0.019,0.013]
[-0.003,0.004]
[-4.807,2.347]
[-1.730,3.045]
[-0.141,0.571]
[-0.372,0.318]
[-0.443,0.343]
[-0.110,0.416]
[-0.071,0.319]
[-0.115,0.140]
[-0.272,0.008]
[-0.029,0.017]
[-0.023,0.013]
[-1.018,0.411]
[-0.013,0.028]
[0.010,0.119]
[-0.055,0.014]
[-0.007,0.000]
[-0.021,0.128]
[-0.136,0.030]
DOCUMENTO DE TRABAJO N.º 1537
[-9.591,20.04]
[1.639,6.182]
[2.841,6.214]
[-0.436,0.064]
[-0.775,-0.251]
[0.092,0.515]
[0.004,0.027]
[-0.107,0.183]
[-0.013,0.010]
[-0.002,0.004]
[-7.473,-0.846]
[-0.244,4.200]
[-0.301,0.264]
[-0.182,0.331]
[-0.419,0.177]
[0.015,0.547]
[-0.144,0.223]
[-0.154,0.075]
[-0.207,0.028]
[-0.033,0.005]
[-0.014,0.020]
[-1.096,0.015]
[-0.019,0.010]
[0.033,0.127]
[-0.049,0.009]
[-0.006,0.000]
[-0.010,0.101]
[-0.107,0.005]
[-1.754,18.03]
[1.242,4.282]
[2.436,4.606]
[-0.339,0.077]
[-0.584,-0.171]
[0.052,0.388]
[0.003,0.023]
[-0.071,0.195]
[-0.005,0.008]
[-0.004,0.000]
[-5.343,-1.122]
[-0.389,2.898]
[-0.253,0.171]
[-0.075,0.235]
[-0.381,-0.031]
[0.072,0.394]
[-0.009,0.226]
[-0.107,0.044]
[-0.188,-0.034]
[-0.020,0.003]
[-0.014,0.008]
[-0.849,-0.169]
[-0.011,0.009]
[0.043,0.109]
[-0.051,-0.009]
[-0.003,-0.001]
[-0.008,0.057]
[-0.073,-0.013]
Table A.2: Posterior mean and 95% probability intervals of the frontier parameter distributions
in profit models
Model P1
No risk covariates
Mean
95% PI
β0
β1
β2
β3
β11
β12
β13
β22
β23
β33
δ1
δ2
δ11
δ12
δ22
η11
η12
η21
η22
η31
η32
κ1
κ2
φ1
φ2
φ3
ϕ1
ϕ2
5.6560
0.0533
0.0927
0.0475
0.0712
0.0186
-0.0048
0.0116
0.0018
0.0011
0.1544
0.1802
0.2195
-0.2212
0.2010
0.1508
-0.0298
-0.0175
-0.0836
0.0012
0.0035
-0.3458
0.0022
0.0364
-0.0344
0.0002
-0.0401
0.0167
BANCO DE ESPAÑA
35
[4.4570, 6.9950]
[0.0023, 0.1381]
[0.0015, 0.2178]
[0.0018, 0.1184]
[0.0192, 0.1282]
[-0.0382, 0.0698]
[-0.0110, 0.0008]
[-0.0350, 0.0591]
[-0.0021, 0.0057]
[-0.0009, 0.0031]
[0.0046, 0.4351]
[0.0065, 0.6177]
[0.0861, 0.3221]
[-0.3014, -0.1322]
[0.0978, 0.3008]
[0.0996, 0.2006]
[-0.0766, 0.0154]
[-0.0744, 0.0267]
[-0.1286, -0.0365]
[-0.0059, 0.0094]
[-0.0040, 0.0103]
[-0.5945, -0.0978]
[-0.0070, 0.0111]
[0.0141, 0.0591]
[-0.0513, -0.0179]
[-0.0013, 0.0018]
[-0.0680, -0.0127]
[-0.0067, 0.0408]
DOCUMENTO DE TRABAJO N.º 1537
Model P2
Common risk coefficients
Mean
95% PI
5.8790
0.0296
0.0792
0.0516
0.0780
0.0119
-0.0044
0.0157
0.0015
0.0010
0.1484
0.1726
0.1866
-0.2062
0.1857
0.1492
-0.0289
-0.0283
-0.0752
0.0017
0.0023
-0.3589
0.0022
0.0373
-0.0349
0.0001
-0.0417
0.0166
[4.9180, 6.8760]
[0.0036, 0.0621]
[0.0070, 0.2172]
[0.0047, 0.1095]
[0.0292, 0.1290]
[-0.0388, 0.0597]
[-0.0095, 0.0007]
[-0.0265, 0.0576]
[-0.0018, 0.0048]
[-0.0007, 0.0027]
[0.0108, 0.3585]
[0.0126, 0.4098]
[0.0581, 0.2884]
[-0.2777, -0.1289]
[0.0980, 0.2720]
[0.1040, 0.1930]
[-0.0680, 0.0113]
[-0.0858, 0.0165]
[-0.1165, -0.0324]
[-0.0042, 0.0080]
[-0.0043, 0.0083]
[-0.5671, -0.1485]
[-0.0054, 0.0096]
[0.0184, 0.0562]
[-0.0489, -0.0209]
[-0.0012, 0.0013]
[-0.0653, -0.0184]
[-0.0029, 0.0364]
Model P3
Random risk coefficients
Mean
95% PI
5.3440
0.0873
0.0401
0.0571
0.0873
0.0022
-0.0029
0.0033
0.0012
0.0014
0.0959
0.1515
0.0397
-0.1467
0.1831
0.1623
-0.0354
-0.0979
-0.0439
0.0033
-0.0016
-0.3092
0.0051
0.0364
-0.0359
-0.0006
-0.0388
0.0099
[4.1010, 6.7390]
[0.0039, 0.1750]
[0.0015, 0.1079]
[0.0024, 0.1810]
[-0.0033, 0.1604]
[-0.0694, 0.0835]
[-0.0099, 0.0034]
[-0.0754, 0.0694]
[-0.0029, 0.0054]
[-0.0003, 0.0032]
[0.0034, 0.2848]
[0.0050, 0.4716]
[-0.1571, 0.2172]
[-0.2555, -0.034]
[0.0751, 0.2814]
[0.0868, 0.2325]
[-0.0929, 0.0279]
[-0.1702, -0.0233]
[-0.1008, 0.0130]
[-0.0035, 0.0117]
[-0.0104, 0.0057]
[-0.5264, -0.0984]
[-0.0025, 0.0125]
[0.0184, 0.0562]
[-0.0489, -0.0209]
[-0.0012, 0.0013]
[-0.0651, -0.0131]
[-0.0105, 0.0314]
BANCO DE ESPAÑA PUBLICATIONS
WORKING PAPERS
1416
DIEGO J. PEDREGAL, JAVIER J. PÉREZ and A. JESÚS SÁNCHEZ-FUENTES: A toolkit to strengthen government
budget surveillance.
1417
J. IGNACIO CONDE-RUIZ, and CLARA I. GONZÁLEZ: From Bismarck to Beveridge: the other pension reform in Spain.
1418 PABLO HERNÁNDEZ DE COS, GERRIT B. KOESTER, ENRIQUE MORAL-BENITO and CHRISTIANE NICKEL:
Signalling fiscal stress in the euro area: a country-specific early warning system.
1419 MIGUEL ALMUNIA and DAVID LÓPEZ-RODRÍGUEZ: Heterogeneous responses to effective tax enforcement:
evidence from Spanish firms.
1420 ALFONSO R. SÁNCHEZ: The automatic adjustment of pension expenditures in Spain: an evaluation of the 2013
pension reform.
1421 JAVIER ANDRÉS, ÓSCAR ARCE and CARLOS THOMAS: Structural reforms in a debt overhang.
1422 LAURA HOSPIDO and ENRIQUE MORAL-BENITO: The public sector wage premium in Spain: evidence from
longitudinal administrative data.
1423 MARÍA DOLORES GADEA-RIVAS, ANA GÓMEZ-LOSCOS and GABRIEL PÉREZ-QUIRÓS: The Two Greatest. Great
Recession vs. Great Moderation.
1424 ENRIQUE MORAL-BENITO and OLIVER ROEHN: The impact of financial (de)regulation on current account balances.
1425 MAXIMO CAMACHO and JAIME MARTINEZ-MARTIN: Real-time forecasting US GDP from small-scale factor models.
1426 ALFREDO MARTÍN OLIVER, SONIA RUANO PARDO and VICENTE SALAS FUMÁS: Productivity and welfare: an
application to the Spanish banking industry.
1427 JAVIER ANDRÉS and PABLO BURRIEL: Inflation dynamics in a model with firm entry and (some) heterogeneity.
1428 CARMEN BROTO and LUIS MOLINA: Sovereign ratings and their asymmetric response to fundamentals.
1429 JUAN ÁNGEL GARCÍA and RICARDO GIMENO: Flight-to-liquidity flows in the euro area sovereign debt crisis.
1430 ANDRÈ LEMELIN, FERNANDO RUBIERA-MOROLLÓN and ANA GÓMEZ-LOSCOS: Measuring urban agglomeration.
A refoundation of the mean city-population size index.
1431
LUIS DÍEZ-CATALÁN and ERNESTO VILLANUEVA: Contract staggering and unemployment during the Great Recession:
evidence from Spain.
1501
1502
LAURA HOSPIDO and EVA MORENO-GALBIS: The Spanish productivity puzzle in the Great Recession.
LAURA HOSPIDO, ERNESTO VILLANUEVA and GEMA ZAMARRO: Finance for all: the impact of financial literacy training
in compulsory secondary education in Spain.
1503
MARIO IZQUIERDO, JUAN F. JIMENO and AITOR LACUESTA: Spain: from immigration to emigration?
1504
PAULINO FONT, MARIO IZQUIERDO and SERGIO PUENTE: Real wage responsiveness to unemployment in Spain:
asymmetries along the business cycle.
1505
JUAN S. MORA-SANGUINETTI and NUNO GAROUPA: Litigation in Spain 2001-2010: Exploring the market
for legal services.
1506
ANDRES ALMAZAN, ALFREDO MARTÍN-OLIVER and JESÚS SAURINA: Securitization and banks’ capital structure.
1507
JUAN F. JIMENO, MARTA MARTÍNEZ-MATUTE and JUAN S. MORA-SANGUINETTI: Employment protection legislation
and labor court activity in Spain.
1508
JOAN PAREDES, JAVIER J. PÉREZ and GABRIEL PEREZ-QUIRÓS: Fiscal targets. A guide to forecasters?
1509
MAXIMO CAMACHO and JAIME MARTINEZ-MARTIN: Monitoring the world business cycle.
1510
JAVIER MENCÍA and ENRIQUE SENTANA: Volatility-related exchange traded assets: an econometric investigation.
1511
PATRICIA GÓMEZ-GONZÁLEZ: Financial innovation in sovereign borrowing and public provision of liquidity.
1512
MIGUEL GARCÍA-POSADA and MARCOS MARCHETTI: The bank lending channel of unconventional monetary policy:
the impact of the VLTROs on credit supply in Spain.
1513
JUAN DE LUCIO, RAÚL MÍNGUEZ, ASIER MINONDO and FRANCISCO REQUENA: Networks and the dynamics of
firms’ export portfolio.
1514
ALFREDO IBÁÑEZ: Default near-the-default-point: the value of and the distance to default.
1515
IVÁN KATARYNIUK and JAVIER VALLÉS: Fiscal consolidation after the Great Recession: the role of composition.
1516
PABLO HERNÁNDEZ DE COS and ENRIQUE MORAL-BENITO: On the predictability of narrative fiscal adjustments.
1517
GALO NUÑO and CARLOS THOMAS: Monetary policy and sovereign debt vulnerability.
1518
CRISTIANA BELU MANESCU and GALO NUÑO: Quantitative effects of the shale oil revolution.
1519
YAEL V. HOCHBERG, CARLOS J. SERRANO and ROSEMARIE H. ZIEDONIS: Patent collateral, investor commitment
and the market for venture lending.
1520
TRINO-MANUEL ÑÍGUEZ, IVAN PAYA, DAVID PEEL and JAVIER PEROTE: Higher-order risk preferences, constant
relative risk aversion and the optimal portfolio allocation.
1521
LILIANA ROJAS-SUÁREZ and JOSÉ MARÍA SERENA: Changes in funding patterns by Latin American banking systems:
how large? how risky?
1522
JUAN F. JIMENO: Long-lasting consequences of the European crisis.
1523
MAXIMO CAMACHO, DANILO LEIVA-LEON and GABRIEL PEREZ-QUIROS: Country shocks, monetary policy
expectations and ECB decisions. A dynamic non-linear approach.
1524
JOSÉ MARÍA SERENA GARRALDA and GARIMA VASISHTHA: What drives bank-intermediated trade finance?
Evidence from cross-country analysis.
1525
GABRIELE FIORENTINI, ALESSANDRO GALESI and ENRIQUE SENTANA: Fast ML estimation of dynamic bifactor
models: an application to European inflation.
1526
YUNUS AKSOY and HENRIQUE S. BASSO: Securitization and asset prices.
1527
MARÍA DOLORES GADEA, ANA GÓMEZ-LOSCOS and GABRIEL PEREZ-QUIROS: The Great Moderation in historical
perspective. Is it that great?
1528
YUNUS AKSOY, HENRIQUE S. BASSO, RON P. SMITH and TOBIAS GRASL: Demographic structure and
macroeconomic trends.
1529
JOSÉ MARÍA CASADO, CRISTINA FERNÁNDEZ and JUAN F. JIMENO: Worker flows in the European Union during
the Great Recession.
1530
CRISTINA FERNÁNDEZ and PILAR GARCÍA PEREA: The impact of the euro on euro area GDP per capita.
1531
IRMA ALONSO ÁLVAREZ: Institutional drivers of capital flows.
1532
PAUL EHLING, MICHAEL GALLMEYER, CHRISTIAN HEYERDAHL-LARSEN and PHILIPP ILLEDITSCH: Disagreement
about inflation and the yield curve.
1533
GALO NUÑO and BENJAMIN MOLL: Controlling a distribution of heterogeneous agents.
1534
TITO BOERI and JUAN F. JIMENO: The unbearable divergence of unemployment in Europe.
1535
OLYMPIA BOVER: Measuring expectations from household surveys: new results on subjective probabilities of future
house prices.
1536
CRISTINA FERNÁNDEZ, AITOR LACUESTA, JOSÉ MANUEL MONTERO and ALBERTO URTASUN: Heterogeneity
of markups at the firm level and changes during the great recession: the case of Spain.
1537
MIGUEL SARMIENTO and JORGE E. GALÁN: The influence of risk-taking on bank efficiency: evidence from Colombia.
Unidad de Servicios Auxiliares
Alcalá, 48 - 28014 Madrid
E-mail: [email protected]
www.bde.es